Telemetry

It is _worse_ than none, because that "some" is non-representative.

No. It's not. I recommend a good primer on statistics. You have to have enough data to draw meaningful statistics from it. Otherwise your data is skewed and worthless, or it's skewed, worthless, and because you don't realize you're working with a non-representative sample, you draw the wrong conclusions from the statistics you do have. I don't have a problem with properly anonymized and useful telemetry gathering in open source when the process can be audited from the beginning to the end. I do have a problem with people making uninformed accusations about projects with telemetry trying to understand how people, in general, use their software.

Telemetry, by the nature of the beast, should be opt-out. That way those that don't mind need do nothing to contribute, while those that don't want to contribute or have a situation where they can't do so may turn it off. The projects will get enough data to draw meaningful conclusions. You don't get that when it's opt-in (tyranny of the default - most people never change them even in Linux land).

Now, saying that, I'm tempering that with meaningful consent. That is, the user is fully informed of what the telemetry is collecting, how its stored and anonymized, and what the methodology and results of the data collection are. This absolutely not the case with proprietary software as there's no meaningful consent in most such schemes. Open source has the opportunity to be both better than that and still demonstrate how telemetry can benefit those projects both in resource allocation from developers, and benefiting the users - with caveats about relying too heavily on statistics for all decisions.